mod_backup
mod_backup serves two different purposes: it makes a nightly backup of your files and database, and can also backup/restore individual resource items.
Daily backup of database and files
Losing data is bad for business. This applies to your customers as well if you are building sites for them. It is critical to keep backups of any Zotonic sites you develop.
After enabling mod_backup, it will make a backup of the site’s data and configuration every night at 3 AM. It keeps the last 7 daily copies of the data and a weekly backup for the last 4 weeks, so you have always a backup to roll back to.
The backups are stored under backup
in the files directory of
your site. Check in the admin under System > Status to see where the
site files directory is located.
The site’s media files are stored as a <site-name>-N.tar.gz
file. The
configuration is named config-<site-name>-N.tar.gz
, while the
database is stored compressed in a <site-name>-N.sql.gz
file. N is
the day number. Sunday (day 1) is stored as the weekly backup, which rotates
over w1..w4.
It is possible to encrypt the backups by enabeling the Encrypt Backups
option on the configuration page of the backup service. When you enable
this for the first time, a password will be generated automatically.
This password should be stored in a safe location to be able to
decrypt the backups. When encryption is enabled, the files have
the .enc
extension.
The zotonic shell command decrypt
can be used to decrypt encrypted
zotonic backup files.
If mod_filestore is enabled then the media files are not backed up, as it is assumed that the files are already backed up on the cloud filestore. The data backups are uploaded to the cloud filestore, so they are also backed up.
If mod_filestore is not enabled then we advise to add a cron script to the server for copying the data to remote storage.
Per-resource revision log
If a resource is updated or deleted then a copy of the resource data is saved to a revision log.
Using the revision log a resource can be rolled back to an older revision or, when deleted, recovered.
Edges are added to the revision log, and are replayed in reverse chronological order to recover the connections of a resource.
If a connection is re-instantiated and it is referring to a dependent resource that is now deleted, then that dependent resource will also be recovered.
Medium files are kept for 4 weeks before they are deleted. That means that up to four weeks a medium file can be recovered, and after that it will be lost. Unless the file is still available on the filestore (which has its own deletion interval).
Revisions are pruned daily and deleted if:
- older than 18 months;
This can be changed by setting the configuration
mod_backup.revision_retention_months
to another number of months. - of user resources and older than 90 days;
This can be changed by setting the configuration
mod_backup.user_revision_retention_days
to another number of days. - of users resources and the users was deleted for at least 30 days;
This can be changed by setting the configuration
mod_backup.user_deletion_retention_days
to another number of days (maximum 30).
Replication and failover
It is possible to replicate data from one server to another server. The receiving server
must have the site running with its environment set to backup
.
If a site’s environment is set to backup
then the site will start in a special mode. In backup
mode only essential modules, mod_filestore
and mod_backup
are started.
If there is a working filestore configuration then the backup module will start polling for new backups uploaded to the filestore. Any newly found backups are downloaded to the local file system. After all backup files are downloaded, then for the newest backup:
- the database backup will be imported; and
- the configuration files in the
priv/
andpriv/config.d/
overwritten with the config files from the backup; and - certificates and secrets are written to the site’s security directory; and
- any files backup tar will be unpacked in the local files directory; and finally
- a file
priv/BACKUP
will be written.
After this the site will restart.
Because the file priv/BACKUP
is present the environment of the site will be forced to backup
and the site will start checking again for a new backup.
If a failover is needed, then change the DNS to point to the new server, remove the priv/BACKUP
file and restart the site to start the site normally.